Docente de pre jardín B

In this section we demonstrate exactly how the model presented in Chapter 3 is used in the scenario. There is a variety of agents in the scenario, but we concentrate on the stages that involve the User Agent, the VO Manager Agent and a number of Service Provider Agents (VO formation stages 4, 5 and 6 from the scenario). More specifically, we use the following agents:

• One User Agent,aua1.

• Four VO manager Agents,avom1,avom2,avom3andavom4.

• Six Service Provider Agents (for simplicity, we assume that each agent in the system has the ability to provide only one service):

– asp4,asp5andasp6providing the HTML content service.

We begin by stating that there is a need to create a VO to meet a specific requirement to provide a composite multimedia communication service to an end user. The composite service consists of the following basic services: HTML content provision and phone calls.

Firstly,aua1 sends the user’s requirements as a query toavom1. Agentavom1 parses the query

and realises that it can only meet these requirements by forming a VO with an agent that can supply a service for phone calls and an agent that can supply a service for HTML content. Agentavom1 is aware of three agents that can provide a phone call service, and its interaction

history with these is shown in Table 4.2. Similarly, it is aware of three agents that are capable of providing HTML content, and its past interactions with these entities are given in Table 4.3.

Agent Past interactions

Successful Unsuccessful

asp1 17 5

asp2 2 15

asp3 18 5

TABLE4.2: Agentavom1’s interaction history with phone call service provider agents.

Agent Past interactions

Successful Unsuccessful

asp4 9 14

asp5 3 0

asp6 18 11

TABLE4.3: Agentavom1’s interaction history with HTML content service provider agents.

Agentavom1 would like to choose the most trustworthy phone call and HTML content service

provider. The following describes how this is achieved using TRAVOS. Before we calculate which of the possible candidates are the most trustworthy, we must specify certain parameters thatavom1 requires. First, we specify the level of error thatavom1 is willing to accept when

determining the confidence in a calculated trust value as = 0.2. Second, we specify that

θγ= 0.95, below whichavom1will seek other opinions about the trustee. 4.2.2.1 Calculating Trust and Confidence

Using the information from Tables 4.2 and 4.3,avom1can determine the number of successful

interactionsn, and the number of unsuccessful interactionsm, for each agent it has interacted with. Feeding these into Equation 3.5,avom1can obtain shape parameters for a beta distribution

function that represents the behaviour of each service provider agent. For example, the shape parametersαandβ, forasp1, are calculated as follows:

Using Equation 3.5: α= 17 + 1 = 18andβ= 5 + 1 = 6.

The shape parameters for each agent are then used in Equation 3.6 to calculate a direct trust value for each agent thatavom1 is assessing. For example, the trust valueτadvom1,asp1 forasp1is calculated as follows:

Using Equation 3.6: τ_ad_vom1,asp1 = _αα_+β = ₁₈₊₆18 = 0.75.

For avom1 to be able to use the trust values it obtains for each agent, it must also determine

the confidence it has in the calculated trust value. This is achieved by using Equation 3.19 and(which is 0.2). For example, the confidenceγavom1,asp1 thatavom1 has in the trust value τ_ad_vom1,asp1 is calculated as shown below:

Using Equation 3.19: γavom1,asp1 = Rτ d avom1,asp1− τd avom1,asp1+ Bα−1(1−B)β−1dB R1 0 Uα−1(1−U)β−1dU = R0.55 0.95 Bα −1_(1−B)β−1_dB R1 0 Uα−1(1−U)β−1dU = 0.98

Agent α β τavom1,ax γavom1,ax

asp1 18 6 0.75 0.98 asp2 3 16 0.16 0.98 asp3 19 6 0.76 0.98 asp4 10 15 0.40 0.97 asp5 4 1 0.8 0.87 asp6 19 12 0.61 0.98

TABLE4.4: Agentavom1’s calculated trust and associated confidence level for HTML content

and phone call service provider agents.

The shape parameters, trust values and associated confidence for each agent,asp1toasp6, which

avom1 computes using TRAVOS, are shown in Table 4.4. From this, it is clear that the trust

values for agents, asp1,asp2 andasp3, all have a confidence aboveθγ (0.95). This means that

avom1 does not need to consider the opinions of others for these three agents. Agentavom1 is

able to determine thatasp3 is the most trustworthy out of the three phone call service provider

agents and chooses it to provide the phone call service for the VO.

4.2.2.2 Calculating Reputation

The process of selecting the most trustworthy HTML content service provider is not as straight- forward. Agentavom1has calculated that out of the possible HTML service providers,asp5has

the highest trust value. However, it has determined that the confidence it is willing to place in this value is 0.87, which is belowθγ and means thatavom1 has not yet interacted withasp5

enough times to calculate a sufficiently confident trust value. In this case,avom1 has to use the

that it can compare to the trust values it has calculated for other HTML providers (asp4 and

asp6).

Let’s assume that avom1 is aware of three agents that have interacted with asp5, denoted by

avom2,avom3andavom4, whose opinions aboutasp5are(15,46),(4,1)and(3,0)respectively2.

Agentavom1can obtain beta shape parameters based solely on the opinions provided, by using

Equations 3.8 and 3.7, as shown below:

Opinions from providers: avom2 = (15,46), avom3= (4,1)andavom4= (3,0)

Using Equation 3.7: N = 15 + 4 + 3 = 22, M = 46 + 1 + 0 = 47

Using Equation 3.8: α= 22 + 1 = 23, β= 47 + 1 = 48

Having obtained the shape parameters, avom1 can obtain a trust value forasp5 using Equation

3.6, as follows:

Using Equation 3.6: τ_ar_vom1,asp5 = _αα_+β = ₂₃₊₄₈23 = 0.32

Now avom1 is able to compare the trust in agents asp4,asp5 andasp6. Before calculating the

trustworthiness ofasp6, agentavom1considersasp5 to be the most trustworthy (see Table 4.4).

Having calculated a new trust value for agent asp5 (which is lower than the first assessment),

agent avom1 now regards asp6 as the most trustworthy. Therefore avom1 chooses asp6 as the

service provider for the HTML content service.

4.2.2.3 Coping With Inaccurate Opinions in the VO

The methodavom1uses to assess the trustworthiness ofasp6, as described in Section 4.2.2.2, is

susceptible to errors caused by reputation providers giving inaccurate information (as discussed in Section 3.4.2). In our scenario, supposeavom2 provides the HTML content service too, and

is in direct competition withasp6. Agent avom1 is not aware of this fact, which makesavom1

unaware thatavom2 may provide inaccurate information aboutasp6 to influence its decision on

which HTML content provider agent to incorporate into the VO. If we look at the opinions pro- vided by agentsavom2,avom3 andavom4, which are(20,46),(4,1)and(3,0)respectively, we

can see that the opinion provided byavom2does not correlate with the other two. Agentsavom3

andavom4 provide a positive opinion of asp6, whereas agentavom2 provides a very negative

opinion. Ifavom2 is providing an inaccurate account of its experiences withasp6, we can use

the mechanism discussed in Section 3.4.2 to allowavom1 to cope with this inaccurate informa-

tion, and arrive at a better decision that is not influenced by self-interested reputation providing agents (such asavom2).

Before we show how TRAVOS can be used to handle inaccurate information, we must assume the following. Agentavom1 obtained reputation information fromavom2, avom3 andavom4 on

2

Opinions can be represented as a pair consisting of the number of successful and unsuccessful interactions the opinion provider says it has had with the trustee.

Agent Weighting Adjusted Values

µ σ α β

avom2 0.0039 0.5 0.29 1.0091 1.0054

avom3 0.78 0.65 0.15 5.8166 3.1839

avom4 0.74 0.62 0.17 4.3348 2.6194

TABLE4.5: Agentavom1’s adjusted values for opinions provided byavom2,avom3andavom4.

[0,0.2] [0.2,0.4] [0.4,0.6] [0.6,0.8] [0.8,1] Total

n m n m n m n m n m

avom2 2 0 11 4 0 0 0 0 2 3 25

avom3 0 2 1 3 0 0 22 10 6 4 30

avom4 1 3 0 2 0 0 18 8 5 3 25

TABLE 4.6: Observations made byavom1given opinion from a reputation source. n repre-

sents that the interaction (to which the opinion applied) was successful, and likewise m means unsuccessful.

several occasions, and each timeavom1 recorded the opinion provided by a reputation provider

and the actual observed outcome (from the interaction with an agent to which the opinion is applied). Each time an opinion is provided, the outcome observed is recorded in the relevant bin. Agentavom1keeps information of like opinions in bins as shown in Table 4.6.

Using the information shown in Table 4.6, agentavom1can calculate the weighting to be applied

to the opinions from the three reputation sources by applying the technique described in Section 3.4.2.3. In so doing, agentavom1uses the information from the bin, which contains the opinion

provided, and integrates the beta distribution between the limits defined by the bin’s boundary. For example,avom2’s opinion falls under the[0.2,0.4]bin. In this bin, agentavom1has recorded

thatn = 15andm = 3. Thesenandm values are used to obtain a beta distribution, using Equations 3.13 and 3.14, which is then integrated between 0.2 and 0.4 to give a weighting of 0.0039 forasp6’s opinion. Then, by using Equations 3.13 and 3.14, agentavom1 can calculate

the adjusted mean and standard deviation of the opinion, which in turn gives the adjustedαand

βparameters for that opinion. The results from these calculations are shown in Table 4.5. Summing the adjusted values for α andβ from Table 4.5, avom1 can obtain a more reliable

value for the trustworthiness ofasp5. Using Equation 3.4,avom1calculates a trust value= 0.62

forasp5. This means that from the possible HTML content providers,avom1 now seesasp5 as

the most trustworthy and selects it to be a partner in the VO. Unlikeavom1’s decision in Section

4.2.2.2 (whenasp6was chosen as the VO partner), here we have shown how a reputation provider

cannot influence the decision made byavom1by providing inaccurate information.